Differences Between Banded Thickets (Tiger Bush) at Two Sites in West Africa

Journal of Vegetation Science II: 321-328, 2000 O IAVS; Opulus Press Uppsala. Printed in Sweden 321 t Differences between banded thickets (tiger bush) at two sites in West Africa

Couteron, Pierre1*, Mahamane, Ali2, Ouedraogo, Paul3 &/ Seghieri, Josiane4

IENGREF, B.P. 5093, F-34033 Montpellier, France; 2Université Abdou Mouinouni, Faculté d'Agronomie, B.P. 10960, Niamey, Niger; 31N.E.R.A, Département Productions Forestières, 03 BP 7047 Ouaga 03, Burkina Faso; 41RD, B.P. 5045, F-34033 Montpellier, France; *Corresponding author; Fcu; i-33467047101; E-mail [email protected]

Abstract. This paper deals with the influence of edaphic condi- oriented parallel to the contour. Banded patterns have also tions on the spatial structure of banded thickets or tiger bush been studied in Mexico (Cornet et al. 1992) and Australia two (Brousse tigrée). It is based on sites in West Africa, with (Mabbutt &Fanning 1987; Ludwig & Tongway 1995). similar climatic conditions but located on contrasting substrates. Vegetated bands accumulate scarce resources (soil, The spatial structure was described with standardized charac- water, nutrients) that run off bare areas (Cornet et al. teristics including thicket spacing, thickethiter-thicket contrast, upslope/downslope asymmetry and species zonation 1992; Ludwig & Tongway 1995; Seghieli et al. 1997). throughout the vegetation band. Recruitment and senescence This process can be analysed in terms of a positive features of woody stands were emphasized in order to under- feedback loop that vegetation may exert on itself in arid stand current dynamics.Data were collected ontransects oriented climates (Wilson & Agnew 1992;Dunkerley 1997). How- perpendicular to the contours and so to the thickets as well. A ever, the consequences of resource re-allocation in terms standardized analytical procedure was applied to data from both of patch structure and subsequent dynamics are less sites to ensure consistent and thorough delineation of thickets. straightforward, even if several studies reported banded The overall periodicity of thickets, the woody flora and the systems wi?: (1) strong contrast between thicket and dominant species Corizbretunz iiiicrarztlzurnwere similar at the inter-thicket; (2) strong asymmetry between upslope and two sites. However, thicket spacing, thickethter-thicket contrast and upslope/downslope floristic asymmetry of the thickets downslopeparts of a thicket; (3) clear ecologicalzonation were higher in the less favourable site. Also seedlings were less with distinct species (andor life stages) occurring along abundant, with a greater dependence on pre-existing thickets. the slope. Not all banded vegetation systems show sharp contrasts From these elements certain papers inferred peculiar and are strongly asymmetric, since such characteristics are vegetation dynamics such as thicket migration against the likely to be reinforced by adverse environmental conditions. slope (Comet et al. 1992; Montaña 1992; Thiéry et al. As a consequence, current dynamics may be more diverse than 1995), whereas other authors remain sceptical (White expected. Quantified inter-site comparisons can greatly help 1970; Wickéns & Collier 1971; Mabbutt & Fanning to classify African banded vegetation systems and to discuss 1987;Couteron et al. 1996) or at least circumspect (López- potential dynamic outcomes. Portillo 1996; López-Portillo & Montaña 1999). Most available field characterizations nevertheless dealt with Keywords: BurkinaFaso; Mortality;Niger; Recruitment; Seed- banded systems that were located under rather arid ling; Spatial pattern. situations - annual rainfall 150-400" and values for ETP (potential evapotranspiration according to Pen- Nomenclature: Hutchinson & Dalziel(1954-1972). man) of ca. 2000 mm - whereas extensive areas of West African tiger bush are encountered when the Abbreviations:BD = site in Burkina Faso; BN = site in Niger; annual rainfall is between 400 and 700 mm, for similar TBA = Total Basal Area; WR =Water Reserve. ETP-values (Hiemaux & Gérard 1999). A more favourable climate may determine different spatial struc- Introduction tures and dynamic outcomes. Furthermore, banded systems were also observed on contrasting substrata In Africa, tiger bush or 'brousse tigrée' has been re- (White 1971; Wickens & Collier 1971; Hiernaux & portedfromMauritaniato Somalia and Sudan (MacFadyen Gérard, 1999). the properties of which may influence 1950;Clos-Arceduc 1956;Boaler & Hodge 1964; Boudet the actual water balance experienced by vegetation. 1972; White 1970; Wickens & Collier 1971; Lawesson The aim of this paper is to compare the spatial 1995; Lepfun 1999). This vegetation type is composed structure of two West African banded vegetation Sys- of altemating, densely vegetated bands and bare areas 322 Couteron, P. et al. flora and human uses, but located on different substrates. 50cm in the inter-band zone; intermediatefigures (1.60m The question addressed is whether contrast, asymmetry and 0.60m) were found, respectively, for the upslope and and species zonation (with related potential dynamics) downslope fringes of the thicket. are influenced by edaphic conditions. A consistent inter- In northern Burkina Faso, tiger bush patterns occur site pattern quantification was attempted in terms of either on old Palaeozoic sandstones or on granites of the overall structure (thicket periodicity and size), thicket Precambrian 'Mossi' shelf (Leprun 1999). We investi- characteristics (contrast and asymmetry) and species gated tiger bush on deep granitic alterite (often more distribution. Special attention was paid to spatial pat- than 2 m), presuming that we would have more favour- terns of seedlings and dead trees, to obtain insights into able edaphic conditions than on the iron-capped pla- present thicket dynamics. teaux of Niger. The site is located on the upper part of a gentle slope (about 0.8%). Soils (10% clay and 70% sand) are poorly developed on the deep sandy-clay Material and Methods alterite, which is easily penetrated by roots. The potential WR of the substratum appeared higher than in BN Study sites (about 25 mm in O - 60 cm layer plus 40 mm in 60 - 130 cm; Guillet 1991), due to a deeper penetrable volume. One site (referred to below as BN) is located in South Moisture within the first 30 cm of the soil was moni- West Niger (13'30' - 13'40' N and 2"40'-2'50'E), near tored gravimetrically during the 1994rainy season. Only Banizoumbouvillage (75 km east of Niamey). The second slight differences were found between the thicket and its site (referredto below as BD) is located in northem Burkina- fringes (Ouedraogo 1997). Faso (around 14"N and 2'30' W), near Bidi village (about 30 km north of Ouahigouya). The climate at both sites is Methods semi-arid tropical with a long dry season (from October to May), alternating with a short rainy season (June to Sep- Observationswere carried out along parallel transects tember). In Niamey the average annual rainfall was 613 2 m wide and 700 m long, with a spacing of 25 m and mm between 1950 and 1967 but decreased with 20% oriented perpendicular to the contours (i.e. a sampling between 1968 and 1985;duringthesameperiodOuahigouya strategy similar to the 'gradsect' approach; Gillison & experienced a 30% decrease (Morel 1992). Brewer 1985). Censuses were performed at the end of The two study sites are located in the 'Sahelian transi- the rainy season (September). The height and the girth tion zone', where most of the woody species are related of the biggest stem (40cm above the ground) were to the 'Sudaniancentre of endemism' (White 1983). Most measured for all woody plants in each of the contiguous of the woody plants were multi-stemmed. In both sites 2m x 2m quadrats. The state (dead or live) of each tree there were no perennial grasses, and the herbaceous cover was noted. The herbaceous cover was visually esti- was composed of annual grasses and forbs. Cyanotis mated in each quadrat (in %). We considered individu- lannata and Microchloa indica were dominant in BN, als of more than 1.5 m in height as mature, i.e. a while Andropogon pseudapricus and M. indica were the threshold that corresponds roughly to the ability for most abundant in BD. Pastoral use was moderate in BN producing diaspores. Additional data on mature indi- and low in BD. Both study sites displayed typical banded viduals of the less frequent woody species were re- patterns with no indication of previous clearing or regular corded in 12 m x 12m quadrats along the same transects woodcutting. to obtain representative samples of their populations. In Niger, the banded pattern occurs only on the laterite- The total basal area of each species was calculated capped plateaux of 'terminal continental' sandstones, directly from the girth for mono-stemmed trees, and which are virtually flat (angle < 0.5% for the study site). through specific regressions based on the girth of the The soils (27% clay and 56% sand) are thin, poorly biggest stem. (r2ranged from 0.7 to 0.8 with respect to developed with petroferric gravels originating from a species, for samples of 194 to 490 individuals.) cemented iron-pan that limit root penetration (Ambouta A standardized procedure for thicket delineation 1984; Seghieri et al. 1997). These soils have a low poten- was necessary to enable consistent inter-site compari- tial Water Reserve (WR) of 17" in the O-6Ocm layer sons. Ludwig & Tongway (1995) used an objective (water content between 3.0 pF units and 4.2 pF units procedure to locate discontinuities (i.e. boundaries) from times soil depth; Hillel 1988). Soil moisture content was transect data. However, in order to cope with low con- monitored over the 1991-1995 period by neutron probe trast bandinter-band patterns (as in BD), we empha- (Galle et al. 1999). A temporary deep drainage occurred sizedpeaks of local density of woody vegetation (i.e. the within thickets every year down to more than a depth of 'core' of thickets). From a data-processingstandpoint, it 5.60m while the infiltration front extendedbarely beyond meant using smoothing filters instead of edge detection ~~

- Differences between banded thickets at two sites in West Africa - 323 ones (Niblack 1986). Thicket identification was based Results on the distribution of mature woody individuals expressed into binary data (occupied quadrats vs. empty), Comparison of the overall woody cover for the sake of robustness. A median filter for binary data was used to identify stretches of the transect with a There were 11 thickets per transect in BN and 12 in high frequency of occupied quadrats, subsequently re- BD. Along the slope, the mean width of thickets was 9.5 ferred as ‘thickets’. (A median filter is well-known as a m (S.D. = 7 m) in BN and 11m (S.D. = 8.5m) in BD. The tool for reducing general noise while preserving the mean width of bare area was 50m (S.D. = 28 m) in BN edges (Niblack 1986: 78)) Filtering was based on a and 41m (S.D. = 21) in BD. Thicket centroids were moving window of five consecutive quadrats (10 m in located 60m apart (S.D. = 28m) in BN and 52m apart the field). For each thicket, the location of the centroid (S.D. = 32m) in BD. Consequently the two sites showed was deduced from the distribution of all individuals a pattern scale of similar size. Combretuin micraiithum found in the constituent quadrats. Quadrats not belong- dominated the woody cover in the two sites (Table l), ing to a thicket were assigned to either the downslope or accounting for 69% of mature individuals in BN (72% upslope margins according to their relative distance to in BD), and for 60% and 57% respectively of the total the nearest thicket edge. A thicket and its two margins basal area (TBA). The mean height of its mature indi- constituted an ‘element’ of the overall banded pattern. viduals (2.4m in BN and 2.6 m in BD) did not differ Margins and thickets were characterized according significantly according to a t-test. to the densities of woody individuals of a given category Woody vegetation was denser in BN than in BD. (species or size class). Results were illustrated by plot- The most abundant species after C. mìcraizthuin were ting elements into an equilateral triangle whose sides Guiei-a senegalensis (13% of TBA) and Coinbretuin expressed the relative proportion of density or cover (%) izigi-icans(14% of TBA) in BN, and Pterocarpus lucens in each of the three components. -This is an analogy of in BD (28% of TBA). Its mature trees were 4.2 m tall the soil texture class triangle.- Highly contrasting ele- on average (S.D. = 2.2 m). Trees taller than 5m were ments with higher vegetation proportions in the thicket scarcer in BN than in BD, accounting respectively for were closer to the upper apex. Asymmetrical elements 0.8% and 17% of the TBA. Also seedlings (below 0.5m with higher proportions in the upslope margin were in height) were less abundant in BN than in BD. In BN, closer to the lower right apex, and vice versa for the 49% of them were C. mici-aiztliumand 43% were Guiera. downslope margin (see Fig. 1). Elements with homoge- InBD 48% were P. lucens and40% were C. inicranthum. neous vegetation distributiontended to be located around Finally, P. lucens was not found at all in BN, although the ‘centre’ of the triangle. For the two dominant species all other species were found in both sites. in each site, emphasis was placed on the density distribution of trees below 0.5m tall, regarded as seedlings. Species distribution within ‘eleineizts’ of the pattern 27 permanent plots (5 m x 5m) were established on each site to assess the survival rate of new seedlings Mature individuals were cross-classified according during their first rainy season. Three thickets were ran- to species and location within the pattern. Some species, domly selected along the transects, with three plots on were too scarce to yield significant results (Fisher’s the upslope margin, three inner plots and three plots on exact test) whatever the site (Cadabafarinosa, Laiznea the downslope margin. New seedlings were tagged and acida, Maerua angoleiisis, Pterocarpus erinaceus). Five registered at the end of one rainy season (in October species yielded significant results on both sites. In BN 1993 in BD and 1994 in BN) from the presence of their (Table 2a), the relative distributionsdiscriminated species cotyledon traces. Plots were revisited at the onset of the that were more frequent in the upslope margin (mainly next rainy season (June). Boscia salicifolia) vs. species that were more frequent For most of the results, two-way contingency tables resulting from cross-classifications of plant numbers Table 1. Main characteristics of woody stands. Specific according to site, height, species, location or condition densities refer to mature individuals. (deadhlive), and sometimes age, were tested for inde- BN BD pendence between rows and columns using ‘Fisher’s Total basal area (m2ha) 1.9 3.1 exact test’ (Sokal & Rohlf 1995). This test is based on Density of mature individuals (Nha) 1565 550 the hypergeometrical distribution and on the number of Density of trees with height > 5 m (Nha) 8 15 413 individuals in the cells of the tables, although it enables Density of Coinbretuin iiiicrantlzzmn (Nha) 1080 Density of Guiera senegaleilsis (Nha) 241 20 indirect comparisons between proportions. Mean values Density of Combretitm izigricaits (Nha) 81

324 Couteron, P. et al.

Table 2. Contingency table of species distributions within contrasted densities between thickets and margins. Fur- elements of the pattem (based on mature individuals). Each thermore, some ‘elements’ had seedlings only in the cell is tested for independencebetween rows and columns (is. margins of the thickets. In BN, Guiera had denser a similar distribution whatever the species) using Fisher’s seedling populations in the upslope margin of most exact test. Symbols + and - express the sign of the departure from the null assumption;p < 0.001 unless otherwise stated elements (Fig. lb), although 9% of them contained no ** =p< 0.01; * =p< 0.05. In bold: species yielding signifi- seedlings. In BD, Guiera seedlings (not plotted) tended cant results on both sites. to be on both margins but their density was low (only 320 individualsha). In BD, seedlings of P. Zzicens were Upslope Downslope margin Thicket margin quite evenly distributed (Fig. lb), with no preferential a. EN site pattern. The herbaceous cover tended to be higher in the Boscia salicifolia + - ** - upper margins in BN, whilst evenly distributed in both Guieru senegalensis + - - margins in BD (Fig. IC). - :k Combretumnigricaiis NS NS To confirm these results, the relative distributions of Coininiphora africana -* + - Acacia inacrostachya NS +* NS seedlings of the dominant species were compared from Boscia angustifolia. - NS NS a contingency table (Table 3). In BN, the relative distri- - Grewiajkvescens + NS bution of C. micranthum seedlings had a significant Combretuni micrantliuiti - i- -k Acucia atmacantha NS + + peak within the thicket and a significant depression Gardenia sokotensis NS +* + outside (P

Banizoumbou site Bid¡ site in both the thicket and the downslope margin (Acacia atmacantha, C. micranthum, Gardenia sokotensis, Grewia Javescens). Only one rather scarce species (Commiphora africana) was clearly more frequent in the thicket than in either margin. In BD (Table 2b), a) Seedlings of ,i/- -y certain species appeared more frequently in the upslope Combretum 100% micranthum 0% 3’ margin, whilst others had a relatively higher frequency 100% oownaiope margin 0% in the thicket. However, most species had similar rela- A n tive frequency in the downslope margin. There was no floristic similarity between thicket and downslope margin, whereas some similarity could be observed be- /\ b) Seedlings ...+.. tween the two margins. of co-dominant woody species Guiera senegalensis Pterocarpus lucens Spatial distribution of seedlings within ‘elements’ of the pattern

Triangles of vegetation proportions were used to compare the distribution of the density of the dominant species seedlings (below 0.5 m) and the herbaceous cover between the two sites. Points concerning the den- Fig. 1. Triangular classification of the elements according to sity of C. micranthum seedlings were concentrated in vegetation distribution in thicket and margins. Overlapping the upper apex of the triangle for both sites (densities elements are indicated by an arrow, the number of elements were higher in the thicket than in the margins; Fig. la). involved is in brackets. Comparison between BN and BD for However, for BD, most points were located at a greater seedlings of dominant species (i.e. C. micuunthum) (a), seed- distance from the upper apex than for BN, denoting less lings of Co-dominant species (b), and herbaceous vegetation (c). - Differences between banded thickets at two sites in West Africa - 325

Table 3. Contingency table of seedling distributions within elements. Each cell is tested for independence between rows and columns (i.e. similar distribution whatever species and site) using Fisher's exact test. Symbols + and - express the sign of the departure from the null assumption (p e 0.001 unless otherwise stated: 'p = 0.24; 2p = 0.41).

In downslope Inside thickets Inside thickets In upslope Total number margins downslope the centroid upslope the centroid margins of seedlings sampled

Combreturn~~ticraiitlium 135 196 186 222 739 (BN) - + + - Guiera sertegalerwis 42 53 192 365 652 (BN) - - + + Contbreturiz micraiitl~um 340 132 96 492 1060 (BD) + -1 +2 Pterocarpus lucens 434 96 60 598 1188 (BD) + - +

57% of the Guiera seedlings were located in an upslope New seedling survival and mortality of mature trees margin against 37% inside a thicket. The relative distribution of P. Zucens seedlings in BD was significantly For each dominant species, Fisher's exact test was higher outside and lower inside the thickets (p

a) Banizoumbou site 140 T 140 8 120 .-c .- Guiera senegalensis p 100 Combretum micranthum $ 80 c ? 60 L 60 R 40 E 40 z '2 20 20 r-l O I L O 4 8 12 16 20 24 28 >28m O 4 8 12 i6 20 24 28 >28m Distance (m) Distance (m)

b) Bid¡ site In downslope margin u In upslope margin

200 t r Combretum micranthum u) t Pterocarpus lucens u) .G 150

2a, 100 2O 5 a 50 zs O 0 4 8 12 16 20 24 28 >28m O 4 8 12 16 20 24 28 >28m Distance (m) Distance (m) Fig. 2. Histograms of the total number of seedlings sampled outside the thickets according to the distance from the nearest thicket edge in BN (a) and in BD (b). Grey = in downslope margins; white = in upslope margins. 326 Couteron, P. et al.

Table 4. Contingency tables for the distribution of Combretian Discussion micruiithuin new germinants after one dry season. Each cell is tested for independence between rows and columns using Zonal similarities inter-site direremes Fisher’s exact test. Symbols + and - = sign of the departure vs. from the null assumption, in bold when Significant. In spite of distinct substrata, the two vegetation Location Total number of Number of systems shared severalbroad characteristics,both floristic new germinants surviving Survival and structural. They were dominated by the same species, tagged germinants rate Combretum micranthum, which is the backbone of these a) Combretum micrarithuin in BN site banded patterns. All other woody species were present Upslope margins 47 8 17% on both sites, except Pterocarpus lucens. However, the - p = 0.013 absence of this species in BN can be related to a Thickets 76 21 36% + p=o.o9 chorological limit, corresponding roughly to the Niger Downslope margins 20 8 40% river, with no obvious ecological cause (Bortoli 1983). f p = 0.22 In BN, thicket spacing was slightly higher than in b) Combretccm micranthum in BD site BD (60 m vs. 52 m). According to field results (White Upslope margins 20 2 10% 1971; Hiernaux & Gérard 1999), and theoretical model- f p = 0.34 ling (Lejeune et al. 1999), there is a decreasing relation- Thickets 62 4 7% - p = 0.60 ship between thicket spacing and mean annual rainfall. Downslope margins 17 O 0% However, the observed spacing values were very simi- - p = 0.31 lar, compared to more arid situations (e.g. 160 m found in Somalia for ca. 150-300 mm of rainfall; MacFadyen 1950). This corroborated the conclusion, deduced from lower in BD than in BN whatever the location. In BD, rainfall and floristic data, that the two sites belonged to the relative distribution of surviving C. micrantlzum the same zonally distributedvegetation type. The shorter displayed no significant difference in relation to loca- spacing in BD was, nevertheless, in agreement with tion (Table 4). The same result was observed for P. other results, such as the greater importance of large lucens (not presented), but this species had a higher trees, and the more extended herbaceous cover, which survival rate than C. micranthum (33% vs. 5%). confirmed more favourable edaphic conditions (deep The same independence test was applied for each of root penetrability and more sandy topsoil). Nonetheless, the two sites to a contingency table that compared the density of woody individuals and basal area were higher distributions of dead and live mature C. micranthum (> in BN, in spite of harsher soil conditions and of a less 1.5m tall). The number of dead trees of the other species extended vegetation cover. These results may be ex- was too low to be analysed. In BN the proportion of plained by the efficiency of a more contrasted pattern to dead mature C. micranthum was significantly higher (p trap scarce resources (Noy-Meir 1973; Ludwig & <0.05) in the downslope margin, with no significant Tongway 1995), with an adaptable balance between the difference between the upslope margin and the thicket. inter-patch area (runoff source) and the patch size/ In BD, the relative distribution of dead individuals was biomass (runoff sink). Hence, not only patch size and significantly depressed within the thicket (p < 0.0 1) and inter-patch distance (White 1971), but also contrast and slightly higher in the upslope margin (p=0.05). At BN, asymmetry may be reinforced by adverse edaphic con- many dead stumps of unidentifiable species were local- ditions. A similar conclusion may be drawn as to homo- ized near the downslope edge of the thickets (visual geneous soils along rainfall gradients, although this has observation),where they constituted ‘a senescence zone’ never explicitly been stated in the literature. sensuAmbouta(1984) andSeghieri et al. (1997). InBD, no similar feature was observed, since dead stumps Spatial dynamics of species were scarce and scattered all over the banded pattern (pers. obs.). In BN, seedlings of C. micranthum were more frequent inside thickets and had rather high survival rates. In BD, seedlings were more frequent outside, but their survival rates were lower than in BN. In the two sites, regeneration from seeds did exist within the thickets at a level that can easily compensate the observed mortality rates. Recruitment may be limited by existing adults in thickets and by low survival rates outside, a kind of intra-specific control that might explain the persistence 1 - Differences between banded thickets at two sites in West Africa - 327 d of a fragmented woody cover (Winkler & Klotz 1997). Thiéry et al. 1995). A light-demanding species (e.g. Seedlings of C. micranthum did not appear along sharp Guiera) located on the upper edge has been supposed to transitions (‘ecotone dependent’) as did those of pave the way to more water-demanding and shade- Flourensia cernua in Mexican banded vegetation tolerant species (as C. nticranthum). However, spatial (Mauchamp et al. 1993). Thicket extension on their zonation may also be interpreted in terms of persistent periphery appeared possible during very favourable niche separation in a previously established pattern - years, as occurred in the early 1990s. But there was no see Keddy (1989) for examples -with minor change i.e. clear asymmetry in favour of the upslope side. In BD, extensiodcontractions, resulting from rainfall fluctua- the trend toward extension was bidirectional and weak, tion (Jeltsch et al. 1997). Results from the BN site did due to low survival rates. In BN, the upslope edge had a not allow such an interpretation to be discarded in lower survival rate than that observed downslope. On favour of upslope migration. In BD, due to the absence both sites, the spatial pattern of dead individuals matched of any obvious slope oriented asymmetry, the pattern is pretty well the distribution of seedlings, suggesting a probably static. Several reasons can explain why woody kind of balance between trends toward extension and patches may not move upslope, such as: (1) greater contraction(dead individualsoccurred frequently in places water infiltration in the core of the thicket than in the where seedlingswere abundant andor had ahigh survival upslope zone (Cornet et al. 1992; Seghieri & Galle rate). All the above observations support a vision of 1999); (2) sufficient water availability without runoff stable, highly persistent, thickets of C. nzicranthurn. compensation,due to high rainfall and favourableedaphic Most individualsbelonging to other important species conditions (as in BD); (3) extension of the rhizosphere such as P. lucerzs and Guiera were located outside the of trees in the upslope zone (argument proposed by thickets, despite a less favourable water budget, This White 1970). Indeed, the size of adult trees may enable distribution denoted lower water requirements than C. them to exert significant influences (favourable/adverse) nzicraìzthunz, and low competitive ability (hence the on both sides of a patch, so ‘trees could have become exclusion from the best places). In BN, Guiera was self-stabilizing’ (Wickens & Collier 1971). Further- restricted to the upper margin, whilst, in BD, P. lucens more, certain theoretical models demonstrated that static was able to establish on both sides of the bands, some- banded systems may exist (Dunkerley 1997; Lejeune & times far from the thickets. This species displayed abun- Tlidi 1999).As a consequence, upslope migration should dant seedlings, high survival rates and few dead trees. not be considered as an intrinsic property of all banded Hence there was a trend toward an increase in density of systems (White 1971;López-Portillo &Montaña 1999), P. lucens all over the area, which did not mean an since alternative dynamics are possible. extension of the thicket itself. Indeed, spatial distribution of mature P. luceris generally displays little aggre- gation (Couteron & Kokou 1997), and closed canopy Acknowledgements.This study has been partially funded by stands have never been observed. Other woody species the ‘Savannas in the Long Term’ (SALT) programme (IGBP/ are too scarce to be thought of as influencing the overall GCTE core project). We are grateful to Pierre Hiernaux (ILRI, vegetation dynamics. Niamey, Niger), Jean Thiéry (CEA, Cadarache, France) and David Tongway (CSIRO, Australia) for their advice.

Potential overddynamics References In BN, seedlings of dominant woody species were highly dependent on the existing thickets, i.e. a strong Ambouta, K. 1984. Contributionà 1‘édaphologie de la brousse patchlinter-patch contrast, but only Guiera showed a tigrée de l’ouest izigéi-ieiz. Thèse de Docteur Ingénieur, clear asymmetry in favour of the upslope side, as did the Université de Nancy I, Nancy. herbaceous cover. Conversely, the distribution of dead Boaler, S.B. & Hodge C.A.H. 1964. Observations on vegeta- individuals proved asymmetric in favour of the down- tion arcs in the northern region, Somali Republic. J. Ecol. slope side. In BD, the contrast was weaker, and no 52: 5 11-544. Bortoli, L. 1983. Pterocarpus lucens en Haute-Volta. In: substantial asymmetry was detected for either seed- les Toutain, B. (ed.) Espèces ligneuses et herbacées dans lings, herbaceous cover or dead individuals. Ecological écosystènm piiturés de Haute-Volta, pp. 57-65. IEMVT, zonation of species was observed on both sites, but this Maisons-Alfort. followed a slope oriented asymmetry in BN whereas it Boudet, G. 1972. Désertificationde l’Afrique tropicale sèche. expressed a thicket vs. margin opposition in BD. Adansonia (Sér. 2) 12: 505-524. A slope-oriented zonation of species within the Clos-Arceduc, M. 1956. Etude sur photographies aériennes banded system was sometimes considered as supporting d’une formation végétale sahélienne: la brousse tigrke. the hypothesis of thicket migration (Cornet et al. 1992; Bull. I. F. A. N. 7-18 (Sér.A): 678-684. ~

328 Couteron, P. et al.

Comet, A., Montaña, C., Delhoume, J.P. & López-Portillo, J. demography of Prosopis glanddosa var. torreyaiia in 1992. Water flows and the dynamics of desert vegetation vegetation arcs and associated bare areas. J. Veg. Sci. 7: stripes. h.Hansen, A.J. &Di Castri, F. (eds.) Landscape 901-910. boundaries: consequences for biotic diversi9 and eco- Ludwig, J.A. & Tongway, D.J. 1995. Spatial organisation of logical jZow~~,pp. 327-345. Springer-Verlag, New York, landscapes and its function in semi-arid woodlands, Aus- NY. tralia. Landscape Ecol. 10: 51-63. Couteron, P. & Kokou, K. 1997. Woody vegetation spatial Mabbutt, J.A. & Fanning, P.C. 1987. Vegetation banding in patterns in a semi-arid savanna of Burkina Faso, West arid Western Australia. J. Arid Environ. 12: 41-59. Africa. Plant Ecol. 132: 21 1-227. MacFadyen, W.A. 1950. Vegetation patterns in the semi-desert Couteron, P., Mahamane, A. & Ouedraogo, P. 1996. Analyse plains of British Somaliland. Geogr. J. 116 199-210. de la structure de peuplementsligneux dans un fourré tigré Mauchamp, A., Montaña, C., Lepart, J. & Rambal, S. 1993. au Nord-Yatenga (Burkina Faso): état actuel et con- Ecotone dependent recruitment of a desert shrub, Flozirensia séquences évolutives. Ann. Sci. For. 53: 867-884. cemua, in vegetation stripes. Oikos 68: 107-116. Dunkerley, D.L. 1997. Banded vegetation: development un- Montaña, C. 1992. The colonization of bare areas in two- der uniform rainfall from a simple cellular automaton phase mosaics of an arid ecosystem. J. Ecol. 80: 315-327. model. Plaizt Ecol. 129: 103-111. Morel, R. 1992.Atlas agro-climatique des pays de la zone du Galle, S., Ehrmann, M. &Peugeot, C. 1999. Water balance on CZLSS. Notice et commentaire. AGRHYMET, Niamey. a banded vegetation pattern. A case study of tiger bush in Niblack, W. 1986. An introduction to digital image process- Western Niger. Catena 37: 197-216. ing. Prentice-Hall, Englewood Cliffs, NJ. Gillison, A.N. & Brewer, K.R.W. 1985. The use of gradient Noy-Meir, I. 1973. Desert ecosystems: environment and pro- directed transects or gradsects in natural resource surveys. ducers. Annu. Rev. Ecol. Syst. 4: 25-5 l. J. Environ. Manage. 20: 103-127. Ouedraogo, P. 1997. Rôle des termites daizs la structure et la Guillet, F. 1991. Etude et modélisation hydrologique d'un dyiiamique d'urze brousse tigrée soudano-sahélienne. bassin versant de la zone sahélo-soudanienne au Burkina Thèse de doctorat, Université de Paris VI, Paris. J. Faso. Thèse de Doctorat, Université de Paris VI, Paris. Seghieri, & Galle, S. 1999. Run-off contribution to a Sahelian Hiernaux, P. & Gérard, B. 1999. The influence of vegetation two-phase mosaic system: Soil water regimes and vegeta- pattern on the productivity, diversity and stability of veg- tion life cycles. Acta Oecol. 20: 209-217. J., etation: The case of the 'brousse tigrée' in the Sahel. Acta Seghieri, Galle, S., Rajot, J.L. & Ehrmann, M. 1997. Oecol. 20: 147-158. Relationships between the soil moisture regime and the Hillel, D. 1988. L'eau et le sol. Principes et processus phy- growth of the herbaceous plants in a natural vegetation siques. Academia edition, Louvain. mosaic in Niger. J. Arid Environ. 36: 87-102. Hutchinson, J. & Dalziel, J.M. 1954-1972. Flora of West Sokal, R.R. & Rohlf, F.J. 1995. Biometry: the principles and Tropical Africa. Crown Publishers, London. practice of statistics in biological research. Freeman, San Jeltsch, F.,Milton, S.J.,Dean, W.R.J. &vanRooyen, N. 1997. Francisco, CA. Simulated pattern formation around artificial waterholes Thiéry, J.M.,

Received 3 November 1997; Revision received 14 January 1998; Final version received 27 August 1999; Accepted 25 September 1999. Coordinating Editor: E. van der Maarel.